METHOD AND APPARATUS FOR IMPROVED SAMPLING RESOLUTION IN X-RAY IMAGING SYSTEMS
The present invention pertains to an apparatus and method for X-ray imaging wherein a radiation source comprising rows of discrete emissive locations can be positioned such that these rows are angularly offset relative to rows of sensing elements on a radiation sensor. A processor can process and allocate responses of the sensing elements in appropriate memory locations given the angular offset between source and sensor. This manner of allocation can include allocating the responses into data rows associated with unique positions along a direction of columns of discrete emissive locations on the source. Mapping coefficients can be determined that map allocated responses into an image plane.
This application is a Continuation Application of the co-pending, commonly-owned U.S. patent application with Ser. No. 13/738,614, U.S. Pat. No. 9,217,719filed Oct. 1, 2013, by A. Lowell et al., and entitled “Method and Apparatus for Improved Sampling Resolution in X-Ray Imaging Systems,” which is hereby incorporated by reference in its entirety.
FIELD OF THE INVENTIONThe present invention pertains to X-ray imaging systems. The present invention pertains more specifically to X-ray imaging systems utilizing sources having a plurality of discrete emissive locations.
BACKGROUNDPoint-source X-ray imaging systems currently account for the greatest portion of medical X-ray imaging systems in the United States. Point-source X-ray imaging systems comprise an X-ray radiation source such as an X-ray tube that emits X-rays from a single discrete location or window. While point-systems have achieved a relatively high level of image resolution and imaging speed, they are limited by an unfavorable signal-to-noise ratio related to the position of the patient relative to the source and detector and provide data sufficient only for flat, e.g. two-dimensional, images.
X-ray imaging systems have been developed that address these latter two deficiencies. See for example U.S. Pat. No. 5,729,584 entitled “Scanning Beam X-Ray Imaging System,” issued to Moorman et al. In contrast to a point-source imaging system, this type of imaging systems utilizes an X-ray source having a plurality of discrete emissive locations on its face and a relatively small detector. The geometry of this type of imaging system both improves signal-to-noise ratios by decreasing the number of scattered X-ray photons collected and provides sufficient data to reconstruct a range of planes between the source and detector.
However, the spatial resolution and image fidelity provided by these systems can vary from plane to plane, and can range from being competitive with or better than advanced point-source systems and being quite poor. Embodiments of the present invention provide a method and apparatus of improving the resolution and image fidelity of such systems.
SUMMARYThe present invention pertains to an apparatus and method for X-ray imaging wherein a radiation source comprising rows of discrete emissive locations can be positioned such that these rows are angularly offset relative to rows of sensing elements on a radiation sensor. The angular offset can be less than 90 degrees or less than 5 degrees, or may displace a sensing element by a length equal to an integer number of sensing elements, including one, two, or three sensing elements. A processor can process and allocate responses of the sensing elements into appropriate memory locations given the angular offset between source and sensor. This manner of allocation can include allocating the responses into data rows associated with unique positions along the direction of columns of discrete emissive locations on the source. Mapping coefficients can be determined that map allocated responses into an image plane. Responses can be mapped along a first dimension, aggregated, and then mapped along the second dimension, or may be mapped directly into an image plane.
These and other objects and advantages of the various embodiments of the present invention will be recognized by those of ordinary skill in the art after reading the following detailed description of the embodiments that are illustrated in the various drawing figures.
The present invention is illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which like reference numerals refer to similar elements.
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings. While the invention will be described in conjunction with these embodiments, it will be understood that they are not intended to limit the invention to these embodiments. On the contrary, the invention is intended to cover alternatives, modifications and equivalents, which may be included within the spirit and scope of the invention as defined by the appended claims. Furthermore, in the following detailed description of embodiments of the present invention, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be recognized by one of ordinary skill in the art that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures, components, and circuits have not been described in detail as not to unnecessarily obscure aspects of the embodiments of the present invention.
U.S. Pat. No. 5,729,584 entitled “Scanning Beam X-Ray Imaging System,” U.S. Pat. No. 6,876,724 entitled “Large-Area Individually Addressable Multi-Beam X-Ray System and Method of Forming the Same,” both of which are hereby incorporated by reference, describe X-ray imaging systems utilizing sources that emit radiation from a plurality of discrete locations on their faces and spatially resolved sensors. These and similar imaging systems can acquire sufficient data to reconstruct multiple image planes between the source and sensor and can be useful for fluoroscopic guidance, image acquisition for computed tomography, and other applications.
Examples of a source capable of emitting radiation from a plurality of discrete locations may be an array of carbon nanotube cathodes or other nanotube cathodes, scanning electron beam sources, scanning laser sources, and arrays of single cathode emitters.
Spatially resolved sensors in imaging systems of the present invention may comprise a sensor array of small sensors or sensor elements, e.g. a pixelated detector. A sensor array may comprise elements or pixels of photon-counting detectors, energy-integrating detectors, energy-resolving detectors, or any type of detector sensitive to X-ray photons. A single sensor element or pixel may not be spatially resolved, e.g. the location within an individual sensor element where a photon strikes may not be recoverable, but a sensor array can be spatially resolved by associating data with the position of the sensor element or pixel from which it is received.
The subsets of discrete source locations and sensor elements which sample a given point in an image plane, e.g. the discrete source locations that emit and sensor elements that receive plurality of rays 103 to sample point 121 in image plane 280, are related to the position of said image plane between the source and sensor array. While plurality of rays 103 can be grouped together to reconstruct point 121 in image plane 280, plurality of rays 103 samples a plurality of different points in subsequent image plane 281. Multiple image planes within the space between a source and sensor can be imaged by processing and combining data according to ray intersections with the respective planes. An individual plane can be selected for viewing, or multiple planes can be combined to provide three-dimensional depth.
A plurality of rays may or may not converge to a point or points in subsequent image plane 281 as rays 103 converged to point 121 in image plane 280. The convergence seen in plane 280 can actually have negative effects for resultant image quality, since rays 103 all sample a single point, providing essentially redundant information, rather than a plurality of points in the image plane which could all contribute meaningful information for the population of image plane with accurate pixel values. The manner in which rays connecting discrete source locations to detector elements converge or fail to converge in various image planes can determine the sampling resolution of the system and thereby the accuracy, spatial resolution or quality of resultant images.
Sampling resolution may be characterized as the average number of points sampled per image pixel. Image pixels can be defined by dividing an image plane into a grid of predetermined dimensions, and taking each square or rectangle of the grid to be an image pixel. Characterizing the average number of sample points, e.g. ray intersections, per image pixel rather than total points per image plane may be preferable, as points may not necessarily be distributed evenly within a plane. Having a high sampling resolution, e.g. a number of sampled points within most or all pixels, can be important for a number of reasons. At a minimum, having a sampling resolution of at least one point per pixel is necessary to assign a measured value to each pixel in an image. Greater sampling resolution can however be necessary to avoid aliasing or other distortions and inaccuracies.
An imaging volume may contain features that are smaller or finer than a single image pixel; regions within an image pixel may provide varying amounts of X-ray attenuation. If an image pixel is sampled at only one or a few points, the value which will be assigned to that pixel may reflect the properties of the single feature or few features within the pixel that are sampled, which may or may not be a good representation of the pixel as a whole. If an image pixel is instead sampled at a large number of points, the pixel can be assigned a value based on the average properties measured throughout the pixel and likely be a more accurate representation. For example, if an image pixel is mostly X-ray transparent but has a relatively small, X-ray opaque feature, the reconstructed pixel may be assigned a very bright value if it happens to be sampled only at the small opaque feature. However, if said pixel were sampled at a number of points throughout its area, it may instead be assigned a dim value, representing the average of a small number of sample points falling on the opaque features and a large number of sample points falling on the X-ray transparent regions.
Other types of aliasing or prealiasing can occur when sampling resolution is low, particularly when an image has repeating, e.g. sinusoidal, features. The Nyquist criterion for signal processing shows that the frequency at which a signal is sampled must be at least twice the frequency of the highest frequency signal component to avoid aliasing, e.g. a type of distortion wherein a high-frequency component is reconstructed as a relatively lower frequency component. Analogously for image processing, a sampling resolution at least twice as narrow as the width of the finest features in an image plane may be desirable to avoid aliasing.
Sampling resolution can affect the spatial resolution and fidelity of reconstructed images. In existing tomosynthetic imaging systems some image planes tend to have poor sampling resolution due to the convergence of rays at a small number of sample points.
Sampling patterns may be simulated by calculating the center of every sensors in a given array, back-projecting rays from each sensor to an array of discrete source locations, and recording ray intersection points in a grid of imaging pixels for a selected imaging plane.
In contrast to the relatively populated sampling pattern and quarter-pixel resolution seen in pixel array 401 of
The presence and locations of ray convergence points can be related to the geometry of a tomosynthetic imaging system, namely, to the relative positions of discrete source locations and sensor elements. Since the sampling pattern in an image plane is created by the intersection of rays connecting discrete source locations with sensor elements, changing position parameters of source locations or sensor elements can change the sampling patterns of each image plane. Referencing
Embodiments of the present invention provide manners of improving the sampling resolution of tomosynthetic imaging systems.
In one embodiment of the present invention, a sensor array is angularly offset such that x- and y-axes of the sensor, defined for the present embodiment as axes to which rows and columns, respectively, of sensor elements run parallel, are not angularly aligned with the x- and y-axes of a source, to which rows and columns, respectively, of discrete source locations run parallel.
An effect of having an angular offset between the source and sensor in this embodiment may be to separate rays which would otherwise converge, e.g. that would sample the same point if the source and sensor were aligned.
Angularly offsetting a sensor array can alleviate sampling redundancy not only in very sparsely sampled planes but also in image planes where any amount of ray convergence exists; offsetting a sensor by a predetermined amount does not merely shift points of convergence among different image planes in the same way that changing the spacing of source locations or sensor elements may but can improve sampling resolution throughout the imaging space of a system.
Angular offsets between a source and sensor array in embodiments of the present invention are not limited to 3 degrees or any other number of degrees. An offset may be any number of degrees between 0 and 360 degrees. Angular offsets in embodiments of the present invention may further be between 0 and 90 degrees, 0 and 45 degrees, 0 and 30 degrees, 0 and 20 degrees, 0 and 10 degrees, or 0 and 5 degrees, inclusive. For example, an angular offset may be 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, or 15 degrees. Angular offsets need not be integer number of degrees and may be rational or irrational fractions of degrees, such as less than 1 degree, between 1 and 2 degrees, between 2 and 3 degrees, between 3 and4 degrees, between 4 and 5 degrees, or any other fractional number of degrees. For example, an angular offset may be any number of degrees, including zero, plus or minus ⅛, 1/7, ⅙, ⅕, ¼, ⅓, ½ a degree, or other fractional or decimal numbers of degrees.
However, there may be angular offsets that do not yield spatial resolution benefits, particularly if the sensor has some amount of rotational symmetry. For example, for a square sensor array, or any other sensor array with 4-fold rotational symmetry, an angular offset of ninety degrees or any multiple of ninety degrees may recreate the same sampling scenario as a non-offset sensor. For a rectangular sensor array, or any other sensor array with 2-fold rotational symmetry, an angular offset of 180 degrees or any multiple of 180 degrees may recreate the same sampling scenario as a non-offset sensor array. Rotating a sensor array by an angle which is rotationally symmetric to a non-angularly offset sensor array, e.g. rotating a sensor array of n-fold rotational symmetry by an angle of 360/n degrees, may not result in spatial resolution benefits.
Amounts of angular offset in embodiments of the present invention may also be characterized as a number of “elements per full length” of the sensor. “Elements per full length” may refer to the fractional or whole number of sensor elements or pixels by which a corner of the sensor is displaced, e.g. vertically or horizontally, from its non-offset position. For example, a counterclockwise sensor angular offset may displace the upper right corner of the sensor array upwards vertically and leftward horizontally. The absolute difference between its initial vertical position and its offset vertical position may be divided by the height of sensor elements to yield a fractional or whole number of sensor elements per full length, or the absolute difference between its initial horizontal position and its offset horizontal position may be divided by the width of a sensor elements to yield a fractional or whole number of sensor elements per full length. This manner of characterization may be particularly useful due to the relationship between sampling patterns and geometries of the source and sensor, specifically, to the size and spacing of discrete emissive locations on a source or elements on a sensor face.
Angular offsets in embodiments of the present invention can be any whole or fractional number of elements per full length, including by less than one element per full length. An optimal degree of sensor angular offset or rotation, e.g. a number of elements per full length that results in the best sampling resolution, may be an integer number of elements per full length of the sensor. For example, in one embodiment of the present invention, a sensor is rotated by one sensor element per full length of the array, and in another embodiment, two elements per full length of the sensor array. Both of these embodiments may exhibit optimized sampling performance. Other integer angular offsets of 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 15, or any other integer number of elements per full length of the sensor can be utilized. Alternatively, a sensor array may be offset by fractional, non-integer numbers of elements per full length, including but not limited to 1/16, ⅛, ¼, ⅓, ½, ⅔, or ¾ of an individual sensor element, or any other fraction of a sensor element less than, greater than, or between the enumerated values.
Small angles of offset, e.g. small numbers of elements per full length, may utilize available x-ray radiation efficiently; unless an x-ray source or the collimator of an x-ray source is designed for a specific degree of sensor array offset, positioning a sensor array out of alignment with the source may result in some sensor elements being moved out of the paths of X-ray beams into space where no X-rays will be received. For small angular offsets, e.g. 1, 2, or 3 elements per full length, the benefits of improved sampling may be greater than these losses. For relatively large angular offsets, the decrease in elements that will be illuminated by X-ray radiation may become more significant.
In another embodiment of the present invention, a sensor is fabricated with vertical and horizontal offsets between sensor elements along rows and columns, respectively, and utilized for imaging in conjunction with a non-offset source. Vertical and horizontal offsets may be less than the length of one sensor element, and the resultant configuration of sensor elements may be similar to that of a square or rectangular sensor that has been angularly offset by some amount. In an alternative embodiment of the present invention, a source is fabricated with vertical and horizontal offsets between discreet emissive locations along rows and columns, respectively, and utilized for imaging in conjunction with a non-offset sensor. Either of these embodiments of the present invention can improve sampling resolution by creating an angular offset between the directions of rows and columns on a source and sensor.
The benefits of maintaining an angular offset between a source and sensor in embodiments of the present invention may be attributable to the creation of a finer sampling “grid” from an increase in the number of points at which projections of the source array of discrete emissive locations and sensor array of elements or pixels intersect. This effect may be visualized by projecting two square grid patterns, representing the centers of elements of the source and sensor arrays, onto a wall or other surface, and sliding one grid horizontally across the other in increments of approximately one grid square. If the intersection points of the two grids are marked at each increment first for the case where the axes of the two grids are parallel and secondly for the case where the grid being slid has been rotated by a small degree with respect to the stationary grid, a larger number of intersection points between the two grids for the angularly offset case is visible, demonstrating an improvement of vertical resolution. The same process may be repeated sliding one grid vertically to demonstrate an improvement in horizontal resolution.
Existing image reconstruction methods may be insufficient to quickly handle data from a system utilizing an angularly offset sensor and realize the image reconstruction-quality benefits of an offset sensor array. Apart from slightly more complex geometry, back-projection of the views taken from an angularly offset sensor array onto a focal-plane image, e.g. the assignment of sensor data to points in an image plane, may not be much more conceptually difficult than back-projection of views from a non-offset sensor; rays may be traced from sensor elements back to discrete source locations and intersection points in a given image plane determined.
In one embodiment of the present invention, images are reconstructed by geometrical back projections of each sensor datum, e.g. the response or value of each sensor element upon illumination by each discrete source location, into the image plane. In this embodiment, the non-parallel geometry of an imaging system with an angular offset between the source and sensor is accounted for during preliminary steps of a ray back-projection method. In this embodiment, inputs such as the locations of sensor elements on a sensor, e.g. in sensor x-y coordinates, and the angular offset between the source and sensor can be utilized to generate a grid representing offset locations of sensor elements, e.g. in source x-y coordinates.
The generation of a coordinate grid representing offset positions of sensor elements can also be completed using inputs such as predetermined sensor element spacing, e.g. center-to-center distance, sensor dimensions, and the angular offset. Appropriate sensor coordinates can be calculated from given inputs in any one of a number of ways, including but not limited to application of trigonometric relations, such as sine and cosine, to the distance of an element from the geometric center of the sensor array and the angular offset; simulation of the offset sensor array and sampling of sensor locations; or any other method. Rays can be back-projected from the coordinate grid, through image planes to discrete source locations, to determine intersection points with the planes. Monte Carlo or analytical methods may or may not be used to simulate interactions of the rays within the imaging volume and calculate probable attenuation properties of the medium based on sensor data in order to assign pixel values within the image planes.
However, efficient implementations of image plane reconstruction that can support real-time image reconstruction may process rows and columns of sensor array data to be reconstructed independently, in successive pipelined stages. One such implementation is described in U.S. Pat. No. 6,178,223 entitled “Image reconstruction method and apparatus” issued to Solomon et al, herein incorporated by reference in its entirety.
Efficient reconstruction methods may process or map data from a sensor data set along one dimension as it is received but process or map the data along the second dimension following a predetermined amount of data accumulation or aggregation. These methods can reduce the amount of memory required for storing unprocessed data, e.g. by processing in one dimension as the data is received, while also reducing the overall amount of processing that occurs, e.g. by strategically aggregating data for further processing.
A set of sensor data can be generated for the firing of each discrete source location. As illustrated in
Hardware, firmware, or software components can be configured to apply appropriate scaling factors, or mapping coefficients, along each row. Sets of mapping coefficients can be calculated in real time based on the position of the plane being reconstructed and passed to row processors or may be stored in the row processors or external memory. One or more of these row processors can be utilized to row process each data set. When the source and sensor are aligned, the same set of mapping coefficients may be applied to every row of a given sensor data set as every row in the data set corresponds to the same set of horizontal coordinates along the sensor.
After row processing, data may be stored in the form of “pseudo-rows,” rows associated with sensor element rows but populated with data that has been mapped into the image plane and thereby allocated into image pixel columns. The number of image pixel columns in a pseudo-row may be fewer than the number of columns in the original sensor data set. Pseudo-rows may later be column-processed to complete mapping of the data into a final image. A pseudo-row or number of pseudo-rows from a given data set may be combined or grouped with pseudo-rows from other data sets prior to column processing. Column processing can comprise similar operations as row-processing, where column processors can map vertical sensor coordinates associated with each pseudo-row into coordinates of the image plane and allocate the pseudo-rows into image pixel rows.
This type of method may be termed “separable reconstruction” because the mapping of rows and columns of sensor data into respective rows and columns of image pixels are separable processes. Namely, a sensor data set can be row-processed, e.g. mapped pseudo-rows comprising image pixel-indexed columns, by applying the same set of mapping coefficients to every row; the row processor does not need to identify which sensor-indexed row it is processing to apply correct horizontal mapping coefficients. Similarly, column processing can map entire pseudo-rows into image pixel-indexed rows, e.g. into a final image, by applying a vertical scaling factor to every element or column across a pseudo-row or rows; the column processor does not need to identify a sensor column from which image pixel-indexed information was derived in order to apply correct vertical mapping coefficients across a pseudo-row.
Positioning a sensor array out of alignment with the array of source locations in embodiments of the present invention may break the pattern regularities, e.g. the scaling factor relationships between elements of a sensor data set and a final image plane, that make separable reconstruction possible. When an angular offset exists between a source and sensor, mapping sensor data along sensor rows into an image plane can require applying both column-specific scaling factors as before but also row-specific offsets. Similarly, mapping sensor data into an image plane along sensor columns can entail applying row-specific scaling factors as well as a column-specific offsets. A reconstruction method following that which was previously described may accurately process rows of the sensor data set by providing a row processor for each row of sensor containing mapping coefficients that comprise both scaling factors and offsets. However, applying correct column-specific offsets during column processing would be impossible since no sense of sensor columns exists once data has been mapped into pseudo-rows. Further embodiments of the present invention provide manners in which the benefits of an angularly offset sensor may be incurred without sacrificing the potential to optimize efficiency in image reconstruction.
In one embodiment of the present invention, information sufficient for independent row and column processing may be maintained for separable reconstruction by altering the structure of the sensor data set prior to or during row or column processing. In this embodiment, rows of a sensor data set can be associated with unique y-coordinates of the source, e.g. relative to a y-axis that is aligned with columns of discrete source locations, rather than with the physical rows of elements along a sensor. These “coordinate-indexed” rows may have empty or non-populated columns; each row may contain data in columns only where a sensor element shares the row's unique y-coordinate. Forming data sets of coordinate-indexed rows in this manner can remove the need for applying column-dependent offsets during columns processing, the offsets being essentially incorporated into the structure of the data set, and thereby enable an implementation of separable reconstruction.
In systems where rows and columns of a source and sensor are angularly aligned rows and columns of a sensor data set may correspond to rows and columns of the sensor array; if the sensor of
The length of coordinate-indexed rows can be constant through a data set. In one embodiment of the present invention, the length of each coordinate-indexed row is equal to the number of columns of sensor elements on the sensor. For example, the data set of
In one embodiment of the present invention, separable reconstruction can be implemented using row processors and column processors. Row processors and column processors can be in hardware, firmware, software, or a combination of any of the three. A row processor may exist for every coordinate-indexed row of a given sensor data set and may contain a set of mapping coefficients which can be applied to sensor element data populating a given coordinate-indexed row to map that data to appropriate image pixel-indexed columns of a pseudo-row. In this embodiment, pseudo-rows may also be coordinate-indexed.
Alternatively, a single row processor, or a number of row processors less than the number of coordinate-indexed rows, may be utilized. As memory buffers, e.g. row buffers, may be used to store data corresponding to a given sensor element and data may be handled in row processors one column, e.g. one element or datum, at a time, a row processor may be switched between buffers and thereby handle data from multiple sensor elements. A single row processor, or a number of row processors less than the number of virtual rows, may contain a set or sets of mapping coefficients which can be applied to sensor element data populating various coordinate-indexed rows and may map these row elements into image pixel-indexed columns, e.g. into pseudo-rows. Alternatively, an additional processor can calculate appropriate mapping coefficients, including but not limited to scaling factors and offsets, that should be applied along a given coordinate-indexed row, and transmit these coefficients to a row processor before or as it processes said coordinate-indexed row.
Once row processors have processed the coordinate-indexed rows of
A significantly greater number of coordinate-indexed rows than sensor-indexed rows may be generated for a given data set. For example, if a sensor array has m physical rows of n physical columns, angularly offsetting it can create n unique y-coordinates per sensor-indexed row, e.g. m*n coordinate-indexed rows. A column processor or processors may therefore need to process n times the number of rows for an offset sensor in this embodiment than for a non-offset sensor. Embodiments of the present invention which have been described are in no way limited to 4×4 element sensors but can be applied to sensors of any number of elements. For example, the numbers of rows or columns of sensor elements, e.g. the values of m and n, may be any numbers between 1 and 1,000. Thus, the number of coordinate-indexed rows can become quite large. In a number of following embodiments of the present invention, predetermined approximations can be made for the x- and y-coordinates of sensor elements in order to increase the number of populated elements per coordinate-indexed row and decrease the number of coordinate-indexed rows to be processed.
In one embodiment determination of mapping coefficients for each sensor datum and the allocation of sensor data into coordinate-indexed rows can be based on an approximation of a sensor as a segmented array wherein adjacent segments have been shifted horizontally and vertically relative to one another. The horizontal and vertical shifts between segments can be related to the amount of angular offset between source and sensor.
While the embodiment of
The amount by which sensor segments are shifted in such an approximation may be related to the size of the angular offset between the source and sensor and with the number of segments being utilized for the approximation. More than or fewer than four segments can also be defined for this type of approximation.
In embodiments of the present invention, angularly offset sensors may be modeled with 4, 9, 16, 25, 36, or 49 segments or any number of segments between 4 and 10,000. The number of segments across a sensor face, e.g. in segment rows, may or may not be equal to the number of segments down a sensor face, e.g. in segment columns. Segments may be square or rectangular. The horizontal and vertical shifts between segments in these approximations, e.g. horizontal shift 804 and vertical shift 803, may be any number or fraction of sensor elements such as ¼ elements, ½ elements, ¾ elements, 1 element, 5/4 elements, 3/2 elements, 7/2 elements, 2 elements, and so forth, and any integer or non-integer number of elements between, above or below the enumerated values.
Step 196 may entail the determination of the distance along the direction of sensor element rows, dx(c, r), and the distance along the direction of sensor element columns, dy(c, r), between the center of each segment relative and the geometric center of the sensor. For example, in
Step 197 may entail the determination, by calculation, measurement, or other method, of the physical displacement along the row, Dx, and along the column, Dy, of a corner of an angularly offset sensor relative to its position when aligned with the source. For example, Dx and Dy are shown in
Step 196 may be performed before, concurrently with, or after step 197 and step 198. In step 199 the displacements of segment centers from the sensor center of a segmented model, dx′(c, r) and dy′(c, r), can be calculated as:
dx′(c, r)=dx(c, r)+Sx*[dy(c, r)÷H]
dy′(c, r)=dy(c, r)+Sy*[dx(c, r)÷H]
with dx(c, r) and dy(c, r) from step 196 and Sx and Sy from step 198.
In the embodiment of
An efficient image reconstruction from sensor element data can then be accomplished in block 193. Coordinate-indexed rows can be processed to form pseudo-rows. If a row processor was created for each coordinate-indexed row in step 204, a processor may be applied to each coordinate-indexed row, as in step 206. In step 206, each coordinate-indexed row may be processed simultaneously, if sufficient computing power is available, or sequentially. Alternatively, if a lesser number of row processors was created in step 204, the processor(s) may be applied to multiple coordinate-indexed rows, e.g. switched between locations in row buffers containing element data, as in step 207. It is possible that block 194 and step 206 may be repeated for sensor data corresponding the each discrete source location prior to step 208, column processing. Once a predetermined number of data sets have been mapped into pseudo-rows, or at any other time, column processors created in step 205 may be applied to complete the separable reconstruction. Indicated by step 209, the image may be displayed, stored, or otherwise utilized.
Block 191, block 192, and block 194 may be executed at any time before, during, or after image data acquisition. Steps 202 through 208 may be repeated a number of times corresponding to the number of discrete source locations, e.g. number of sensor data sets, that contribute to a final image. Block 193 and block 194 may also be repeated in order to reconstruct multiple planes, e.g. slices at different distances between the source and sensor. Planes may be displayed independently or overlaid for depth. Blocks and steps in the embodiment of
The parameters and calculations of block 191 may also be considered in the selection of an optimal amount of angular offset to utilize between a source and sensor. For example, an angular offset may be utilized for which the quantities Dx and Dy, and thus Sx, Sy, dx′(c, r), and dy′(c, r) are rational. Embodiments of the present invention comprising a sensor offset by an integer number of elements per length, e.g. one, two, or three sensor elements per length, may cause the above quantities to be rational; a sensor may be offset such that Dx or Dy equals one, two, three, or another integer or rational number of sensor elements.
Approximations or models other than the segmented approach that has been described can also be utilized to decrease processing time in embodiments of the present invention. Such approximations or models may include but are not limited to rounding, grouping, or otherwise manipulating the unique y-coordinates of sensor elements across a sensor row in embodiments of the present invention in a predetermined fashion to decrease the number of coordinate-indexed rows to be processed.
Increasing the size of a sensor, e.g. increasing the numbers of rows and columns of a sensor array, can increase the field of view of an imaging system. However, increasing the number of sensor elements can also increase the processing demands of the system, particularly in embodiments of the present invention wherein m*n coordinate-indexed rows can be created during reconstruction. A rectangular sensor may be utilized to provide a relatively long field of view in one direction and a reasonable but smaller field of view in the other, which can be useful for a variety of fluoroscopic procedures. The number of coordinate-indexed rows created for reconstruction can be limited by segment approximations as previously described.
Both of the embodiments of
An implementation of separable reconstruction from the two-tile sensor configuration of the embodiment of
In one embodiment of the present invention, the two-tile sensor of
Embodiments of the present invention may comprise tomosynthetic imaging systems utilizing rotated or angularly offset, multi-element sensors of any shape or size. Sensors may, but need not, comprise multiple smaller sensor arrays or tiles is in the embodiments of
Embodiments of the present invention comprising real-time row-processing and aggregated column processing may be particularly efficient for an illumination pattern that runs across rows of discrete source locations. However, the direction of the scan may be reversed; source locations along a column may illuminate the image space sequentially, e.g. top to bottom, before restarting at the beginning of the next column. In this case, it may be more efficient to process column data first, e.g. during the scan, implementing coordinate-indexed columns and column processors, and handle row processing after a predetermined number of data sets have been column-processed. While specific embodiments of the present invention may refer to a particular order of row and column processing, the order may be reversed without requiring significant changes to the method described. Furthermore, the definitions of rows and columns need not be strictly associated with horizontal or vertical directions. For example, rows and columns may be related to the pattern in which source locations illuminate the image space, the order in which data is received from the sensor, or a variety of other system parameters.
Reconstruction techniques described above are representative only and do not cover all alternatives techniques or details within. Additional reconstruction aspects, described for a non-offset sensor but which may be utilized, can be found in U.S. Pat. No. 6,178,223 issued Jan. 23, 2001, entitled “Image Reconstruction Method and Apparatus,” U.S. Pat. No. 5,644,612 issued Jul. 1, 1997 entitled “Image Reconstruction Methods,” U.S. Pat. No. 5,751,785 issued May 12, 1998 entitled “Image Reconstruction Methods,” U.S. Pat. No. 6,181,764 issued Jan. 30, 2001 entitled “Image Reconstruction for Wide Depth of Field Images,” all of which are herein incorporated by reference.
The foregoing descriptions of specific embodiments of the present invention have been presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise forms disclosed, and many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and its practical application, to thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the claims appended hereto and their equivalents.
Claims
1. An X-ray imaging system comprising:
- a radiation source configured to emit radiation from a plurality of discrete emissive locations through an imaging volume, wherein said plurality of discrete emissive locations includes rows of discrete emissive locations;
- a radiation sensor having rows of sensing elements, wherein an angular offset exists between said rows of sensing elements and said rows of discrete emissive locations; and
- a processor configured to process and allocate responses of said sensing elements to said radiation from one of said discrete emissive locations into an appropriate memory location.
2. The X-ray imaging system of claim 1 wherein said angular offset is less than 90 degrees.
3. The X-ray imaging system of claim 1 wherein said angular offset is less than 5 degrees.
4. The X-ray imaging system of claim 1 wherein said radiation sensor further comprises columns of said sensing elements.
5. The X-ray imaging system of claim 4 wherein said angular offset is configured to displace one of said sensing elements a distance equal to a length of an integer number of said sensing elements along direction of said columns of sensing elements.
6. The X-ray imaging system of claim 5 wherein said integer number of sensing elements is one.
7. The X-ray imaging system of claim 5 wherein said integer number of sensing elements is two.
8. The X-ray imaging system of claim 5 wherein said integer number of sensing elements is three.
9. A method for image reconstruction comprising:
- positioning an object for imaging between a radiation source having rows and columns of discrete emissive locations and a radiation sensor having rows and columns of sensing elements, wherein an angular offset exists between said rows and columns of sensing elements and said rows and columns of said discrete emissive locations, respectively;
- illuminating a portion of said object with radiation from one of said discrete emissive locations;
- detecting responses of said sensing elements to said radiation;
- allocating said responses into a data set wherein data rows correspond to unique positions along a direction of columns of said discrete emissive locations; and
- determining mapping coefficients relating said data set to an image plane.
10. The method of claim 9 further comprising:
- mapping said data sets into image plane rows along a first dimension.
11. The method of claim 10 further comprising:
- illuminating additional portions of said object sequentially by additional discrete emissive locations;
- detecting additional responses from said sensing elements;
- allocating said additional responses from said sensing elements into additional data sets;
- mapping said additional data sets into said image plane rows along said first dimension;
- aggregating a predetermined number of said image plane rows; and
- mapping said image plane rows into image plane columns along a second dimension.
12. The method of claim 11 further comprising:
- modeling said rows and columns of sensing elements as a collection of rectangular segments shifted by predetermined non-zero shift values relative to one another.
13. A method for optimizing sampling resolution of an imaging system comprising:
- positioning a radiation source that has columns of discrete emissive locations on a first side of an imaging volume;
- positioning a radiation sensor that has columns of sensing elements on a second side of said imaging volume such that an angular offset less than 90 degrees exists between said columns of said discrete emissive locations and said columns of sensing elements;
- illuminating a portion of said imaging volume with radiation from one of said discrete emissive locations; and
- recording responses of each of said sensing elements.
14. The method of claim 13 wherein said angular offset is less than 5 degrees.
15. The method of claim 13 wherein said angular offset is configured to displace one of said sensing elements by a distance equal to a length of an integer number of sensing elements along direction of said columns of discrete emissive locations.
16. The method of claim 15 wherein said integer number of sensing elements is one.
17. The method of claim 13 further comprising:
- sequentially illuminating additional portions of said imaging volume by said discrete emissive locations;
- recording additional responses of each of said sensing elements; and
- reconstructing an image plane from said responses and said additional responses.
18. The method of claim 17 further comprising:
- mapping said responses and said additional responses into said image plane in a first dimension as said responses and said additional responses are received.
19. The method of claim 18 further comprising:
- aggregating said responses and said additional responses mapped in said first dimension into subsets; and
- mapping said subsets into said image plane in a second dimension
20. The method of claim 17 further comprising:
- mapping said responses and said additional responses into said image plane as said responses and said additional responses are received.
Type: Application
Filed: Dec 21, 2015
Publication Date: May 5, 2016
Patent Grant number: 9733198
Inventors: Augustus Percival Lowell (Durham, NH), Tobias Funk (Martinez, CA), Chwen-Yuan Ku (San Jose, CA), Josh Star-Lack (Palo Alto, CA), Edward Gerald Solomon (Melno Park, CA), Winston Y. Sun (Palo Alto, CA)
Application Number: 14/977,279